Fake News

Detection Using
Machine Learning

Submitted By
Krish Goyal
Divyansh Gupta
 Introduction
● In today's information landscape, the
surge in fake news undermines media
credibility and distorts information
dissemination.

● This project is dedicated exclusively to

the utilization of Machine Learning (ML)
for constructing a robust Fake News
Detection System. With a focused
emphasis on ML algorithms, such as
Decision Tree, Naive Bayes, and
Logistic Regression, the aim is to
develop an efficient tool capable of
identifying deceptive narratives.
 The increasing prevalence of misinformation
had seen one such case where numerous
media outlets reported baseless claims
suggesting that COVID-19 was deliberately
engineered as a bioweapon.

1. Undermines public trust.

2. Fuels panic and fear.
3. Distorts factual understanding.
4. Hinders informed decision-making.
5. Promotes misinformation.
6. Erodes credibility of authoritative sources.
7. Incites social unrest.
8. Damages individual and collective well-being.
9. Impedes effective crisis response.
10. Foments division and polarization.
 LITERATURE REVIEW

● Aphiwongsophon and Chongstitvatana [10] presented "Detecting Fake News with

Machine Learning Method" at the 2018 International Conference on Electrical
Engineering/Electronics, Computer, Telecommunications, and Information Technology
(ECTI-CON) in Chiang Rai, Thailand. Their research focused on employing machine
learning methods to identify and tackle the issue of fake news, offering potential
solutions within the technological sphere.

● Gupta, Batla, Kumar, and Jain [8] conducted an in-depth analysis centered around the
comparative evaluation of machine learning models for fake news classification. Their
meticulous study provides valuable insights into the performance, strengths, and
weaknesses of various machine learning techniques employed to discern between
genuine and deceptive news content. Through this comprehensive comparative
analysis, the researchers aim to contribute to the advancement of methods capable of
effectively identifying and combating the dissemination of false information in digital
FALSE NEWS
DATASET

TRUE NEWS
DATASET
 ● Enhance existing fake news
OBJECTIVES detection model performance by
exploring advanced methodologies.

● Expand dataset diversity to enhance

model generalizability across news
domains, languages, and cultures.

● Apply methods to reduce bias for a

fair and ethical use of fake news
detection model.

● Optimize scalability and efficiency of

models for real-time processing in
diverse digital platforms.
METHODOLOGY
 Methodology
● Data Collection and Preprocessing:
Obtained diverse news datasets,
handled missing values, added class
labels, and cleaned text using Pandas
and wordopt() function.

● Feature Engineering: Transformed

text into numerical TF-IDF format,
splitting data for training and testing
in Scikit-Learn.

● Model Training and Evaluation:

Employed Logistic Regression,
Decision Tree, and Naive Bayes
models for classification, evaluating
performance metrics via Scikit-Learn.

● Manual Testing: Developed

manual_testing() function enabling
real-time assessment of news article
classification using trained models.
MANUAL TESTING
OUTPUT
Model
Evaluation
Result Analysis
 RESEARCH GAPS

● Lack of comprehensive analysis across

diverse news domains for ML models in
fake news detection.

● Need for models capable of dynamically

adapting to evolving misinformation
trends in real-time.

● Imperative analysis of biases within ML

models for ethical and fair fake news
detection.

● Limited discussions on computational

scalability and real-time processing
demands in existing studies.
 Future Work
● Model Enhancement Strategies: Exploring ways to make
models better by combining different predictions
through ensemble methods. Improving how we choose
important features to enhance performance.

● Scalability and Real-time Analysis: Creating models that

can handle fast real-time analysis for changing news and
bigger datasets.

● Ethical Deployment and Collaboration: Fixing biases in

models, using them ethically, and working together to
include diverse data for better adaptability in different
news areas and languages.

● User Accessibility and Interface: Making interfaces easy

for users to check if news is credible, so more people can
use them easily..
 Conclusion
● Our research highlights how machine learning
plays a crucial part in fighting misinformation
through automated fake news detection.

● Although impressive accuracies have been

reached, the difficulty of telling real from
misleading content reveals the demand for
even more advanced models.

● The ongoing challenge underscores the need

for sophisticated models that can navigate
the complexities of deceptive digital content.